(Інститут інформаційних технологій та систем НАН України, Видавничий дім "Академперіодика" НАН України, 2025) Музичук, М. А.; Заболотня, Тетяна Миколаївна
The proposed multilevel method for classifying Ukrainian texts by functional style combines statistical analysis, keyword analysis, and contextual analysis based on the BERT model, which accounts for semantic and contextual dependencies in the text. The results support the hypothesis that combining contextual features (generated by BERT) with statistical style parameters yields the highest classifi cation accuracy. This highlights the advantage of the proposed model for tasks requiring high precision and stability in identifying functional text styles.